Efficient multi-organ segmentation in multi-view endoscopic videos using pre-operative priors

Area of Science:

• Medical Imaging
• Robotic Surgery
• Computer-Aided Surgery

Background:

• Minimally invasive robotic surgery benefits from fused pre-operative (pre-op) and intraoperative (intra-op) imaging.
• Accurate segmentation of anatomical structures during surgery is crucial for improved decision-making.

Purpose of the Study:

• To develop an efficient technique for segmenting multiple objects in intra-op multi-view endoscopic videos.
• To leverage pre-op data priors to enhance intra-op segmentation accuracy.
• To enable real-time performance for practical application in robotic surgery.

Main Methods:

• Formulating segmentation as finding 3D pose and non-rigid tissue deformations from 2D intra-op visual cues, guided by 3D pre-op data.
• Developing a closed-form solution incorporating laparoscopic camera motion models.
• Real-time implementation on a single core CPU.

Main Results:

• Demonstrated robustness in segmenting stereo endoscopic videos.
• Validated on both ex vivo and in vivo clinical data from laparoscopic partial nephrectomy.
• Achieved real-time performance, suitable for resource-constrained robotic surgery systems.

Conclusions:

• The proposed method efficiently integrates pre-op and intra-op data for enhanced surgical visualization.
• This technique offers a practical solution for real-time object segmentation in robotic surgery.
• The approach improves surgical decision-making by providing accurate intra-operative anatomical context.